Set up recording state management

[fbraza]

Summary

Introduce the recording state management layer that powers the voice-to-text experience.

Requirements

• Define the RecordingState enum with idle, recording, paused, and stopped.
• Implement a RecordingProvider (ChangeNotifier) that stores the current state, elapsed duration, transcribed text, and waveform data.
• Provide methods (startRecording, pauseRecording, stopRecording, updateWaveform, updateTranscription) that update state and notify listeners.
• Expose computed getters for UI consumption and wire the provider into the voice-to-text screen via ChangeNotifierProvider.
• Add basic unit tests covering state transitions and duration updates.

Acceptance Criteria

• Provider transitions through all states correctly and notifies listeners.
• Duration increments while recording and resets when stopped.
• UI widgets listen to provider updates for transcript, timer, and waveform data.
• Tests verify start/pause/stop flows and waveform/transcription updates.